CA1233925A - Digital loudspeaking telephone - Google Patents

Abstract

ABSTRACT OF DISCLOSURE

A digital loudspeaking telephone for connection to a telephone line, comprised of a microphone, speaker, codec and digital signal processor. Incoming and outgoing signals are received from remote and local subscribers respectively and in response, incoming and outgoing envelope and noise floor signal levels are generated and compared, and one of the incoming or outgoing signals is suppressed. Automatic gain control of outgoing signals is implemented in order to allow smooth switching of the outgoing signals and to suppress background noise from being transmitted by the microphone. The present invention also detects room echo and reverberation, and eliminates spurious channel switching due to echo signals in response thereto.

Description

0~

01 This invent1on rela-tes to telephone 02 sys-tems in general, and in particular to a digital 03 loudspeaking -telephone for use in conjunction with a 0~ digi-tal PAsX.
05 A loudspeaking telephone enables a local 06 telephone subscriber to listen and speak to a remote 07 party without holding a handse-t to his ear. The 08 loudspeaking telephone incorporates a microphone for 09 receiving voice signals from -the local subscriber, and a loudspeaker for reproducing voice signals received 11 from the remote party via a -telephone line. Thus, a 12 normal conversation can be carried on between the two 13 par-ties while the subscriber's hands are otherwise l occupied.
A common problem in many prior art 16 loudspea]cing telephones is feedback. Simultaneous use 17 of the speaker and microphone can result in a positive 18 feedback path being established between -the speaker 19 and microphone and having a closed loop gain of greater than one, resulting in an unstable system.
21 one type of prior ar-t analog loudspeaking 22 telephone utilizes voice operated switching for 23 enabling one of the microphone or speaker in response 24 to detection and comparison of the relative amplitudes of signals carried by microphone and speaker channels, 26 respectively. The other one of the microphone or 27 speaker is disabled so as to eliminate or "break" the 28 Eeedback path and thereby ensure system stability.
29 A disadvan-tage of the aoremen-tioned prior art loudspeaking telephone is that, in the event -the 31 remote subscriber is talking, -the local loudspeaking 32 -telephone -typically switches into speaker mode (i.e., 33 the microphone is disabled). Hence, the remo-te party 3~ is given the impression that the line has gone "dead", which has been determined to be an undesirable effect.
3G A second -type of prior ar-t loudspeaking 37 telephone is described in Canadian paten-t application ~233~

01 serial number 403,582 ox MITEL CORPO~ATIO~, -Eiled on 02 May 21, 1982, and utilizes automatic gain control 03 circuitry in order to attenuate signals on microphone 04 and speaker channels in response to -the relative 05 amplikudes of the signals.
06 In bo-th prior art loudspeaking telephones, 07 voice signals received from the remote party are 08 compared with a predetermined threshold value and 09 transmi-tted through the speaker in the event they are :LO yreater -than the threshold value. However, voice 11 si.gnals are typically at-tenuated duriny transmission 12 ove:r long distance trunks. Consequently the 13 prede-termined threshold value is typically made quite 14 small in order tha-t the speaker will be enabled in response to detection of received voice signals of 16 relatively low amplitude. As a result, noise carried 17 by the long distance trunks, resulting from dialling, 18 cross talk etc., which are typically of greater 19 amplitude than the threshold value, can result in inadvertent spurious channel switchingO
21 Outgoing signals received by microphones 22 of the prior art loudspeaking telephones are 23 transmitted to the remote party in the event their 24 amplitudes are greater than a further threshold value. Because the acoustical environment in which a 26 loudspeaking telephone is uti].ized can vary from one 27 location to another (for instance a noisy factory 28 floor or a quiet office), the aEorementioned further 29 threshold value is required to be adjusted in order that noise signals from the acoustical environmen-t do 31 not cause the microphone to switch on, resulting in 32 the noise signals being transmitted to -the remote 33 party. Also, temperature drift of circuit components 3~ with -time can result in variations in the threshold levels, requiring adjustment which typically requires 36 the services of a trained qualified technician.
37 A further disadvantage oE prior art analog ~2~

0] loudspeaking telephones is that voice signals prom the 02 remote party, which are amplified and broadcast 03 through -the speaker, occasionally bounce off walls or 04 o-ther acoustic reflectors in the local subscriber's 05 acoustical environment. I'hese echo signals are 06 received by the microphone and may be erroneously 07 detected as being originating signals from the local 08 subscriber, causing the microphone -to switch on and Og -the speaker to switch off, and thereby preventing subsequen-t reception ox voice signals Erom -the remo-te :Ll par-ty.
12 Also, in the event incoming voice signals :L3 received from the line and outgoing signals received 14 by the microphone are of approximately equal amplitude, spurious unstable switching can occur 16 between the microphone and speaker channels ox the 17 prior art loudspeaking telephones.
18 Another prior art loudspeaking telephone 19 utilizes separate microphone and speaker housings in an attempt -to overcome the problems associated with 21 voice operated switching The separation between the 22 microphone and speaker effectively reduces the loop 23 gain to less than unity, thereby rendering the system 24 stable. Al-though a more normal type ox conversation can be carried on with this apparatus, the two 26 housings and accompanying wires result in increased 27 complexity and cost. Furthermore, the local 28 subscriber often attemp-ts to speak directly at the 29 louclspeaker rather than into the microphone, thus reducing the amplitude and clarity of the signal 3:L -transmi-tted to the remote party.
32 The present invention uses voice operated 33 digital switching to ensure that only one ox the 34 m:icrophone or speaker is opera-tional a-t a given time, thus eliminating the feedback pa-th. Hence, the 36 microphone and speaker can be disposed in a single 37 housing, thereby overcoming the disadvan-tages oE the

2~J~3 01 la-tter mentioned prior ar-t loudspeaking telephone.
02 In the present invention, digital envelope 03 representations of outgoing and incoming voice signals 04 are yenerated in response to receiving outgoing and 05 incoming voice signals from the microphone and -the 06 -telephone line respectively. Digi-tal represen-tations 07 oE-the background noise present in the acoustical 08 environmen-t of the local subscriber and on the 09 telephone line are also genera-ted and compared with L0 the generated voice signal envelope representations in LL order to distinguish ac-tual speech energy from noise 12 present :Ln -the microphone and speaker channels L3 respectively, (i.e., the "effective" speech envelope 14 signal is de-tected).
According to the present invention, the 16 background noise is continuously monitored in order 17 that changes in the level thereof are automatically 18 compensated for. Thus, the dlgital loudspeaking 19 telephone of the present invention is self-adaptive, such that switching occurs in response to detec-tion of 21 "actual speech" or voice signals, regardless of 22 background noise.
23 pence, the present invention overcomes the 24 disadvantage of the first mentioned prior art loudspeaking telephone wherein a further threshold 26 level is required to be adjusted by a qualified 27 technician in order tha-t the loudspeaking -telephone 28 can function properly in a particular local acoustical 29 environment, such as a factory or office.
The attack time of the generated speech 3l envelope :is made less -than the decay time. Hence, in 32 the even-t -the local subscriber pauses temporaril.y 33 between spoken words, and low amplitude noise signals

3~ are present on the line, -the m:icrophone channel of the local subscriber maintains control since the amplitude 36 oE the local subscriber's generated speech envelope 37 decays slowly -to an amplitude less than -the amplitude 3~

~3~

01 of the noise signals on the line. I~hus, the prior art 02 disadvantage of spurious channel switching is 03 overcome.
04 Also, the present invention implements an 05 automa-tic gain con-trol function with respect to 06 signals on the microphone channel, for suppressing 07 background noise from the local acoustical 08 environment. For example, in the event the local 09 subscriber pauses while speaking, background noise is initially transmitted to the remote par-ty but is LL gradually at-tenuated such -that the remote party does L2 not ge-t the impression tha-t the line has suddenly gone L3 deacl due -to abrupt channel switching. The gain 14 control function is implemented with a quick attack and slow decay, such that the beginnings of words are 16 no-t truncated in the event the local subscriber 17 con-tinues speaking. Hence, the at-tenuation of room 18 noise is gradual as a result of the slow decay time 19 period.
The present invention also comprises 21 circuitry for estimating the expected echo or return 22 loss level and generating a signal indicative 23 thereof. The effective speech envelope signal is 24 compared with the expected echo or return loss signal level, and erroneous channel switching due to 26 recep-tion of echo signals is thereby prevented.
27 The digital loudspeaking~ telephone of the 28 present inven-ti.on direc-tly processes PCM bytes. Thus, 29 advantage can be taken of -the special features oEEered by presen-t day digital PABXs. For instance, a digital 31 signal processor por-tion of -the presen-t invention can 32 be mu].tiplexed among a plurality of subscribers' 33 loudspeaking telephone sets and line or trunk 34 circuits. Hence, the digital signal processor can be disposed on a main con-trol board oE -the PABX, and 36 accordingly each of the subscribers' loudspeaking 37 telephones can be simply comprised oE a microphone, a Ol speaker and codec for converting the PCM bytes -to 02 analog signals, and vice versa. Consiclerable cost and 03 space savings are enjoyed over the prior art 04 loudspeaking telephones as a result of multiplexing 05 the digital signal processor.
06 In general, the invention is a 07 loudspeaking telephone for connection to incoming and 0~ ou-tgoing unidirectional data lines carrying digital 09 represen-tations o:E incoming and ou-tgoing analog L0 signals respectively, comprising a microphone for :1.L transmi-t-ting -the outgoing analog signal, a speaker for 12 receiving -the incoming analog signal, a codec :L3 connected to the microphone and speaker, for receiving L4 the outgoing analog signal and generating the digital representation of the outgoing analog signal in 16 response thereto, and for receiving the digital 17 representation of the incoming analog signal and 18 generating the incoming analog signal in response l9 thereto, a digital signal processor, including circuitry for receiving current incoming and outgoing 21 signal samples of the incoming and outgoing signals 22 from the data lines, and generating incoming and 23 outgoing envelope signal samples and incoming and 24 outgoing noise floor signal samples in response thereto, circuitry for detecting which of a previous 26 one of the incoming or outgoing signals has been 27 suppressed relative to the other, and fur-ther 28 circuitry or comparing one of the incoming or 2g outgoing noise Eloor samples -to a corresponding one of -the incoming or outgoing envelope signal samples in 3:L event the other one of the previous incoming or 32 outgoing signals was suppressed, and suppressinc3 the 33 other one oE the current incoming or outgoing si.gnal 34 samples in the event the corresponcling envelope signal sample is grea-ter than ths corresponding noise floor 36 signal sample.
37 More particularly, the invention is a 3~3 - 6 -01 digital loudspeaking telephone for connection to 02 incoming and outgoing unidirectional data lines 03 carrying digital representa-tions ox incoming and 04 outgoing analog signals respectively, comprising a 05 microphone for transmi-t-ting the outgoing analog 06 signal, a speaker for receiving the incoming analog 07 signal, a codec connected to the microphone and 0~ speaker :Eor receiving the outyoing analog signal and 09 generatiny the cligital representa-tion of the outgoing L0 analog signal in response thereto, and for receiving :I.L the digital representa-tion of the incoming analog 12 signal and generating the incoming analog signal in 13 response thereto. The invention is further comprised 14 of a digital signal processor connected to the lS incoming and outgoing data lines and the codec, for:
16 (1) s-toring one or more predetermined threshold 17 signals (2) receiving -the incoming and outgoing 18 digital signal representations and generating incoming 19 and outgoing digital envelope signa]. representations and incoming and outgoing digital noise floor signal 21 representations respectively, in response thereto, (3) 22 comparing a predetermined one of the incoming or 23 outgoing digital envelope signal representations with 24 a corresponding one of the incoming or outgoing noise floor signal representations, and (4) suppressing -the 26 o-ther one of the incoming or outgoing digital signal 27 representa-tions in the even-t the aforemen-tioned one of 28 the incoming or outgoing digital envelope signal 29 represen-tations is grea-ter than the corresponding one o:E -the incoming or outgoing noise floor signal 31 represen-tations, or (5) summing a predetermined one o:E
32 -the -threshold signals with the other one of the 33 incoming or outgoing noise floor signal 3~ represen-tations in the event the aforementioned one of the s.ignal representations is less than the 36 corresponding one of the noise floor signal 37 representations and generating a surn signal in f r~3 ^l response thereto, and comparing the other one o-E the 02 incoming or outgoing envelope signal representations 03 with the sum signal and suppressing the other one of 04 -the digital signal represen-tations in the event the 05 other one of the envelope signal representa-tions is 06 less than the surn signal, and suppressing the 07 aforemen-tioned one of the digi-tal signal 08 represen-tations in the event the o-ther one oE the 09 envelope signal represen-tations is greater than the 10 sum signal.
LL The invention is also a method of 12 suppressing one of an incoming or outgoing digital 13 signal on incorning and outgoing data lines oE a 14 loudspeaking telephone, comprising the s-teps of receiving current incoming and outgoing signal samples 16 of the incoming and outgoing signals from -the data 17 lines, and generating incoming and outgoing envelope 18 signal samples and incoming and outgoing noise floor 19 signal samples in response thereto, detec-ting which of a previous one ox the incoming or outgoing signals 21 has been suppressed relative to the other, and 22 compaxing one of the incoming or outgoing noise floor 23 samples -to a corresponding one of the incoming or 24 outgoing envelope signal samples in the event the other one of the previous incoming or outgoing signal 26 samples was suppressed, and suppressing the o-ther one 27 of the curren-t incoming or outgoing signal samples in 28 event the corresponding one of the envelope signal 29 sample is grea-ter than -the corresponding noise floor signal sample.
31 For the purposes of clescribing operation 32 of the present invention, reference will be made below 33 -to microphone or speaker channels as being "in 34 control", by which is meant one of the microphone or speaker of the loudspeaking telephone is enablecl, and 36 the other one of the microphone or speaker is 37 disab:Led.

~3~

01 A bet-ter understanding of the invention 02 will be obtained by re-ference to the de-tailed 03 description below, in conjunction with the following 04 drawings, in which:
05 Figure lA is a system block diagram o-f -the 06 inven-tlon in lts broadest form, 07 Figure lB is a system block diagram 08 illustra-ting the inven-tion connected to a bidirectional Og balanced telephone line, :L0 Figure lC is a system block diagram 11. il:Lustratlng a dlgital signal processor portion of the 12 invention connected to a digital switching ne-twork, L3 Figure lD ls a graph illustra-ting analog 14 to digital signal conversion according to law code, Figure 2 is a detailed block diagram 16 showing the principal features of the digital signal 17 processor portion of the present invention, 18 Figure 3 is a schematic diagram of data 19 storage and manipulation circuitry of a preferred embodiment of the digital signal processor portion of 21 the invention, 22 Figure 4 is a schematic diagram of timing 23 and control circuitry of the preferred embodiment of 24 the digital signal processor portion of the inven-tion, and 26 Figure 5 is a block diagram of a gain 27 control circuit of the presen-t inven-tion.
28 Turning now to Figure lA, illustrating a 29 system block diagrarn of -the invention, a microphone 10 is connec-ted to an amplifier 11 :Eor amplifying an 3:L outgoing analog signal. Amplifier 11 is connected to 32 an analog input of a codec 12 for converting -the 33 outgo.iny analog signal in-to a pulse code rnodulated 34 (PCM) signal representation -thereof. The outgoing PCM
signal PCMo, is applied to unidirectional data line 13 36 and thereErom to digital signal processor 14 for 37 processing the signal in order to effect -the required 38 _ 9 _ 33~?~

01 channel switching, echo cancellation, etc., described 02 below with reference to Fiyure 2. The processed 03 outgoing signal PCMol, is then applied to a 04 unidirectional data line 15 on a PCM signal 05 transmission path.
06 A gain control circuit 16A is connected to 07 the data line 15 and to the digital signal processor 08 14 for electing -the aforementioned microphone channel 09 gain con-trol. In operation, gain con-trol circuit 16A
:L0 receives the processecl digital outgoing signal PCMo' ll and a precletermined control signal from the digital L2 si.gnal processor 14, and performs gain adjustment of L3 the signal by conversion to 13 bit linear code and i suhsequent shifting operations, described in greater detail below with reference to Figure 5. The gain 16 adjusted signal PCMo' is reapplied to data line 15 for 17 ~ur-ther transmission.
18 A-n incoming digital signal PCMi, from a 19 unidirectional data line 16, is received by digital signal processor 14, applied to the incoming channel 21 thereof and processed therein as described below with 22 reference to Figure 2. The resulting processed 23 incoming digital signal PCMi', is then applied to 24 codec 12 via a unidirectional data line 17. Codec 12 generates an incoming analog signal representation of 26 the incoming digital signal PCMi', and applies it to 27 an ampliEier 18 for amplification and broadcast 28 through speaker 19.
29 Digital signal processor 14 detects which one oE the incoming or outgoing PCM channels was in 31 control on the previous sample, and whether amplitude 32 parameters oE generated envelope signal samples of -the 33 curren-t sasnples of PCMi or PCMo are greater than 3~ respective noisa Eloor signal samples, echo signal samples and predetermined threshold signal samples, 36 and causes switching of the microphone and speaker 37 channels in response thereto.

2 4'3 01 In addi-tion, digital signal processor 1 02 may be used to generate programmable tones such as 03 ringing and busy tones which are transmi-tted to 04 speaker 19, as described in Eurther detail belowl with 05 reEerence to Figures 3 and 4.
06 Turning now to figure lB, illustra-ting the 07 inven-tion connected to a balanced 2-wire -telephone 08 line via tip and ring leads 20; digital signal 09 processor l is connected via unidirectional data ]ines 15 and 16 to a second codec 21 which is 11 connected via unbalanced output and input leads 22A
12 and 22B respectively, to an analog trunk circuit 23 L3 for connec-tion to the tip and ring leads T and R, 20.
l on outgoing signal from the microphone 10 is conver-ted to a digital signal PCMo, in codec 12.
16 The PCMo signal is applied to DSP 14 for performing 17 the aforementioned channel switching, etc., and 18 an outgoing signal PCMol, is generated in response 19 thereto. The PCMo' signal is gain adjusted in gain control circuit 16A, reapplied to data line lS, 21 converted into an outgoing analog signal in codec 21, 22 and is transmitted to analog trunk circuit 23 for 23 further transmission along the tip and ring leads 20, 24 to a remote central office, etc.
An incoming signal on the tip and ring 26 leads T and R, 20 is applied to analog -trunk circuit 27 23. Trunk circuit 23 implements various impedance 28 matching conversions and transmits the incoming signal 29 to the unbalanced input lead 22B, in a well known manner. The incoming signal is then receivecl from -the 3:L lead 22B by codec 21 and conver-ted therein to an 32 incoming digital signal PCMi, and subsequen-~ly applied 33 to d:igital signal processor 14 for channel switching, 3~ e-tc. The processed incoming digi-tal s:ignal PCMi' is then -transmitted to codec 12 for conversion -to analog, 36 and therefrom to the speaker 19.
37 Turning now to Figure lC, a plurality of ~3~

Ol unidirectional data lines 24 of the PCM signal 02 transmission path, are shown connected to a dlgital 03 switching ne-twork 25 for selecting PCM slgnals from 04 two pairs of the unidirectional da-ta lines Each of 05 the data lines 2~ can be connected to further digital 06 switching networks 25 or codecs 12 or 21. The digital 07 swi-tching network 25 is well known in -the art, and 08 e:Efec-tively mul.-tiplexes the digital signal processor 09 Lo between a plural:ity of local loudspeaking L0 telephones and outside lines, as discussed above.
I:L In operation, a Eirst predetermined pair :L2 of unldirectional data lines is selected by digital 13 switching ne-twork 25 in response to receiving a l predeterrnined control signal CTRL from an external :L5 controller, such as a microprocessor, (not shown).
16 Signals carried by the unidirectional data lines are :L7 applied to digital signal processor 14 in order to 18 effect the aforementioned channel switching, etc. The 19 selected data lines from the first pair are typically connected to unidirec-tional data lines such as lines 21 15 and 16 in Figure lA. The second pair of selected 22 da-ta lines are typically connected to unidirectionaL
23 data lines 13 and 17. Processed signals PCMol and 2~ PCMi' are received from DSP 14 and are applied -to digital switching ne-twork 25 for connection to the 26 second prede-termined pair of unidirectional data lines 27 2~.
28 BeEore proceeding fur-ther, reference will 29 be made brieEly to Figure lD, in order to describe the well known PCM law encoding technique utilized in 3l -the present inven-tion.
32 PCM signals are frequently compressed for 33 t:ransmiss:i.on along -the da-ta lines, such as 15, 16, 13, 3~ :L7 and 2~, e-tc. In North America, digital compression is lmplemented according -to what is commonly known in 36 the ar-t as law code, whi.le in Europe cornpression is 37 effec-ted according -to A-law code In law code the ?

^l mos-t significant bit of a PCM signal sample is a sign 02 bit, wherein a "1" indicates a positive amplitude and 03 a "0" indicates a negative amplitude. With reference 04 to -the graph, in Figure lD, analog signal amplitudes 05 are rneasured along the horizontal axis and PCM signal 06 sample values are measured along -the vertical axis. A
07 first plurality of less significant bits represents a 08 predetermined amplitude segment or chord of the 09 siynaL. tree chords or segmen-ts are illustrated in l E'igure lD by -the reference numeral 26. A second LL plurality of leas-t significant bi-ts represents a L2 discrete amplitude step within the segmen-t or chord.
L3 A series of discrete amplitude steps is illustrated by l the reference numeral 27, in Figure lD. For example, an 8 bit law code typically consists of a mos-t 16 significant sign bit, 3 bits indicative of the chord 17 and bits corresponding -to a particular one of 16 18 amplitude steps within the predetermined chord.
19 Analog signals are converted to law PCM signals, and vice versa, in codec 12 shown in Figure lA, and 21 codecs 12 and 21 in figure lB, in a well known manner.
22 Turning now to Figure 2, a detailed block 23 diagram is shown illustrating the digital signal 24 processor 14 (Figures lA, lB and lC) o -the invention. Incoming and outgoing PCM signals, PCMi 26 and PCMo, are carried by unidirectional data lines 16 27 and l3 respectively, shown connected to PCM buffers 28 200. PCM buffers 200 are connected to the PCM IN and 29 PCM OUT -terminals of an accumulator 201. Processed incoming and ou-tgoing PCM signals PCMi' and PCM~', are 3:L transmittecl to unidirectional data lines 17 and 15, 32 respectively. Data lines 16, 13, 17 and 15 correspond 33 to -the similarly numbered data lines illustrated in 3~ Figures lA, lB and lC. Incoming and outgoing PCM
sl.gnals PCMi and PCMot are received by accumulator 201 36 under control of timing and control circuitry 202 37 connec-ted to control inputs CTRL of the accumulator 201 '~2~

01 via a control bus 203. Output QR of accumulator 201 is 02 connec-ted to the Sl input oE an arithmetic logic unit, 03 AL,U 204, described in further detail below.
0~ Microphone envelope de-tec-t register 205 and 05 line envelope detect register 206 are storage registers 06 for temporarily storing digital sample values 07 correspondiny to -the amplitude envelopes of signals on 08 the microphone and speaker channels, respectively, as Og described in grea-ter de-tail below.
Microphone noise detect register 207 and :LI line noise detec-t register 208 are s-torage registers for 12 storiny digital sample values corresponding to the 13 amplitude envelopes oE noise signals carried by the l microphone and line channels, respectively. Hence, -the stored noise -floor sample value represents the average 16 ambient noise amplitude associated with a particular one 17 oE -the microphone or speaker channels. For instance, 18 signals received from the remote party tend to exhibit 19 higher average ambient noise due to crosstalk and 60 Hz interference from hlgh tension wires, etc., and signals 21 received from -the local subscriber exhibit noise signals 22 due to ambient room noise.
23 Threshold register 209 stores the 24 predetermined digital threshold values for comparison with the envelope signal sample values stored in 26 registers 205 and 206. A series of operations 27 (including the aforementioned comparison), are performed 28 by ALU 20~ on the digi-tal threshold values and -the 29 digita] values stored in registers 205, 206, 207 and 208, described in greater de-tail below. These 3L operations ultimately resul-t in microphone and speaker 32 channel switching.
33 RUMP register 210 is a storage register for 3~ storing a value indicative of the expected room echo (when the speaker channel is în control) or -the 36 predicted return loss for -the microphone (when the 37 microphone is in control). The RAMP register 210 is 6~ r~'~.J~

l utilized to preven-t Ealse channel switching due -to 02 echo signals or room reverberation, as described in 03 further detail below.
0~ Registers 211, 212 and 21~ are s-torage 05 regis-ters -for temporarily storing the results o-E
06 predetermined ones oE the aEorementioned operations, 07 described below in grea-ter detail.
08 The output OUT, oE ALU 204 is connected to 09 the input IN, o-f accumulator 201 and to respective L0 inputs SI, oE registers 205 to 214. Serial outputs S0 lL oE regis-ters 205 to 21~ are connec-ted together and to l2 the S2 input of ALU 204.
L3 Timing and con-trol circuitry 202 is l connec-ted to a plurality of control inputs of LL~ accumulator 201, PCM buffers 200 and ALU 20~, and to 17 enable inputs, E, of regis-ters 205 -to 214 via con-trol 18 bus 203.
19 Control flip-flop 216 is a one bit memory for storing a digital signal indicative of one of the 21 line or microphone channels having been previously in 22 control during processing of the previous PCM signal 23 sample.
24 Modern day digital PABXs typically employ a time slot scheme for transmitting and receiving PCM
26 signals. According to the preferred embodimen-t as 27 designed for use with -the Mitel ST-BUS scheme, a 28 "frame" of digital signals typically consists of 32 29 8-bit PCM "time slots". Incoming and outgoing signal samples (PCMi and PCMo) are typically transmi-tted 3L during prede-termined ones of the 32 time slots.
32 While reference wa5 made above to PCMi, 33 PCMi', PCMo and PCMol as representing signals 3~ comprised of sequences oE PCM samples, reEerence to these terms in the following description is meant to 37 * ST-BIJS is a registered trademark of Mi-tel Corp.
3~3 - 15 -Ol designate individual samples of the PCM signals.
02 In operation, wi-th reference to Figure 2 03 and the flowchart in appendix A, a microphone signal 04 sample PCMol -transmitted during a predetermined time 05 slot oE a first frame (denoted as -the odd frame), is 06 received from the local subscriber's microphone via 07 data line 13, stored in PCM buffers 200, and 08 subsequently loaded into accumulator 210 under control 09 oE t:iming and control circuitry 202, pursuant to step L0 lA oE the Elowchart.
Ll The received microphone signal sample PCMo 12 ls in law code representation, as described above.
L3 In linear code, a "l" in the most significant bit of a sample indicates a negative value. While full conversion oE PCM signal samples to linear code is 16 unnecessary in order to perform arithmetic operations 17 thereon, the most significant bit (i.e., the sign 18 bit) thereof is reset to "0" in order to perform 2's l9 compliment arithmetic in ALU 204. Pursuant to step l of the flowchart, the signal sample PCMo is rectified 21 in accumulator 201 by clearing the sign bit under 22 control of timing and control circuitry 202.
23 Microphone peak detect register 205 24 contains a digital sample value representing the instantaneous amplitude oE the envelope of the signal 26 on the microphone channel duriny -the previous sample, 27 deno-ted as Pun 28 The rectified microphone signal sample, 29 denoted as RECTu(n), is applied to inpu-t Sl of ALU
204. Pu(n-l), is applied from register 205 to input 3L S2 oE ALU 204, under control of timing and con-trol 32 circui-try 202. Pu(n-l) is subtracted Erom RECTu(n) in 33 AL[J 20~, by means oE a 2's compliment addi-tlon. The 3~ result of the sub-traction RECTu(n)-Pu(n-l) is applied to the input IN of accumulator 201. In -the event the 36 resul-t of the subtraction is negative, indicated by a 37 logic "l" in the mos-t signiEicant bit, the microphone ~3~

01 signal amplltude is decaying and the result 02 (RECTu(n)-Pu(n-l)) is then shifted to the right by 8 03 bits in accumula-tor 201 under control of timing and 0~ control circuitry 202. Shifting the resul-t to the 05 right by 8 bits corresponds to an amplitude division 06 by 256. In the event the subtraction in ALU 204 07 yields a "0" in the mos-t signiEicant bi-t, indicating a 08 rising mlcrophone signal ampli-tude, the result 0~ (RECTu(n) - Pu(n-l)), is shifted -to the righ-t by 5 L0 bits in accumulator 201 (corresponding to division by l1 32).
L2 The shiEted result is -then applied to the l Sl input of ALU 204 and added -therein to the previous 14 envelope sample value Pu(n-l). The resul-t of -this addition is stored in register 205.
16 The resulting surn in regis-ter 205 is the 17 curren-t sample value of the microphone envelope 18 signal, Pu(n) and can be represented by the formula:
19 Pu(n) = P~1(n-l) + [RECTu(n) - Pu(n-l)] / 32, in the event the microphone signal amplitude is 21 rising, or 22 Pu(n) = Pun RECTu(n) - Punt / 256, 23 in the event the outgoing signal amplitude is 24 decaying.
The envelope signal generated from 26 successive sample values of Pu(n), approximately 27 tracks the envelope of the microphone signal, and has 28 modera-te attack and slow decay times.
2~ Pursuan-t to step 2 of the flowchart, the current rnicrophone envelope sample value Pu(n), is 31 shiEted into accumula-tor 201. The previous microphone 32 noise floor sample value stored in regis-ter 207, 33 Nu(n-l), is subtracted Erom Pu(n) in ALU 20~. In the 34 event the resul-t oE this subtraction is positive, Nu(n-l) is lncremented by one in ALU 20~, -thereby 36 forming the current microphone noise floor sample 37 value Nun) which is subsequently applied to

4~

01 microphone noise detect register 207. In the event 02 the result of the aforementioned subtraction i.s 03 negative, the previous noise floor value is 0~ decremented by eight in FLU 204.
05 By incrementing the noise floor value by 06 one and decremen-ting by eigh-t in response to the 07 result o:E the aforemen-tioned subtraction, the noise 08 detest regis-ters 207 and 208 func-tion essentially as 09 negative peak detectors for envelope detect registers 10 205 and 206 respectively, (ie., registers 207 and 208 :l.1 exhibit long attacks and short decays).
12 With reference to step 3 oE the Elowchar-t, 13 a curren-t sample of the incoming (or line) signal PCM
l is loaded into PCM buffers 200 and applied therefrom 15 to accumulator 201. Signal sample PCMi is rectified 16 in accumulator 201, as described above with reference 17 to the microphone channel, resulting in the current 18 rectified line signal sample denoted as RECTA,. A
l9 previous sample value of the line envelope signal 20 PL(n-l), s-tored in line envelope detect register 21 206, is subtracted from RECTL(n) in ALU 204, and 22 shifted either 5 or 8 bits to the right under control 23 of timing and control circuitry 202 as described 2~ above, with respec-t to step l. The resul-t is added to 25 PL(n-l) and stored in line envelope detect register 26 ~06.
27 The curren-t sample value of the line 28 envelope signal PL(n) can be represented by the 29 formula:
pL(n)=pL(n-l)-~cREcTL(n)-pL(n-l)] / 32, 31 in the even-t -the incoming signal amplitude is rising, 32 or 33 pL(n)-pL(n-l)-~[REc~L(n)-pL~(n-l)]/~56, 34 in the event the incoming signal amplitude is 35 decaying.
36 Wi-th reference to step 4 of the flowchart, 37 the current line noise floor sample value, ~L(n),is 38 - ~.8 -rJt ~t5 01 calculated and stored in line noise detect regi.ster 02 208 in a similar manner to the calculation of the 03 microohone noise Eloor sample value, described above 04 with reference to s-tep 2 of the flowchart.
05 The noise floor values generated from 06 successive samples of Nu(n) and NL(n) represent the 07 average ambient noise amplitudes on the microphone and 0~ line channels respectively, have long a-ttack and decay 09 times, and consequently do not react to spurious peak L0 envelope signal sample values on the microphone or :L:L line channels.
L2 with reEerence ko step 5 of -the flowchart, 13 the contents of RAMP register 210 are loaded into l ALU 20~ and decremented therein. The decremented .L5 value of RAMP is -then stored in RAMP reyis-ter 210, as 16 discussed in greater detail below.
17 Referring again to step lA, timing and 1i3 control circuitry 202 determines whether the current 19 PCM frame is odd or even. In the event the current 20 frame is even, step 6 is executed.
21 Operations described below, with reference 22 to steps 6 to 14 of the flowchart, are performed on 23 the current sample values s-tored in registers 205 to 24 214. Hence, the postscript "(n)" has been omitted 25 from -the flowchart and the following description.
26 An indication of which of the microphone 27 or line channels was most active on -the previous 2~ sample, vie. which channel was in control, is stored 29 in con-trol flip-flop 216. In a successful pro-totype 30 oE the invention, a "0" stored in flip-flop 216 was 3l indicative of the speaker channel having been 32 previously in control and a "1" stored therein was 33 indicative of the microphone channel having been in 3~ control on the previous sample.
the value of the bit s-tored in flip-flop 36 216 is detected under control of timing and control 37 circuit 202 in order to determine which channel was in 3~ - 19 -, j ~,~3 ~@~

01 control on the previous sample, corresponding to step 02 6 o:E the flow chart in appendix A, 03 In the event the microphone channel was in 04 control, with reference to step 7A of the flowchart, 05 the microphone noise floor sample value Nu stored in 06 register 208 is subtracted in ALU 204 Erom the 07 microphone envelope sarnple value Put The resul-ting 08 effective microphone signal sample value is stored in 09 register 211.
:L0 In the event the result of -the subtraction l:L in step 7A was positive, indicating tha-t -the local i subscriber is still speaking, the microphone channel :1.3 re-tains control and the resul-t stored in register 211 14 is loaded into accumulator 201 and multiplied by 8 (ie. shiE-ted 3 bits to -the leEt) therein. The value 16 stored in RAMP register 210 is then subtracted from 17 8(PU-Nu~ in ALU 20~ and the result is stored in 18 register 214, pursuant to step 8A.
19 In the event the value stored in the RUMP
register is less than 8~PU - Nut, RAMP register 210 is 21 loaded with the value 8(PU - Nu). Hence, the value 22 stored in the RAMP register 210 follows or "tracks"
23 the peaks of the signals on the microphone channel.
24 Considering step 5 again, the contents of RAMP
register 210 are decremented every al-ternate frame, 26 hence -the value stored in RAMP register 210 decays 27 slowly to zero when there is no speech on the 28 microphone channel.
29 Pursuant to s-tep 10, the value (Pu - Nu) s-tored in register 211 is compared to a speech 31 threshold value Tu. If the result of -the comparison 32 i9 posi-tive, voice energy is deemed to be present on 33 the microphone channel and timing and control 3~ circu:it:ry 202 generates a predetermined control signal on bus 203, for storage in a gain con-trol flip-flop 36 described below with reference to Figure 5, pursuant 37 to step llA.

Jo 5 01 In the event the outgoing signal was 02 previously attenuated, the gain control circui-t 16A
03 (Figures lA, B and C) increases the outgoing signal 0~ gain af-ter 256 frames. In the event the comparison of 05 s-tep 10 yields a negative resul-t, the gain of the 06 outgoing signals on -the microphone channel is caused 07 -to be decreased af-ter 2,0~8 frames, according -to step 08 llB. Four gain steps are employed in the preferred 09 ernbodiment of -the invention; -18db, -12db, -6db and LO Odb. I'he gain is adjusted by receiving the microphone 1I sample value PCMol from accumula-tor 201, converting -to 12 ]inear code, shifting a predetermined number of bi-ts L3 to the right, and reconverting to law code within l gain control circuit 16A, and subsequently restoring the shifted sample value in accumulator 201, as L6 cliscussed below with respect to Figure 5.
17 Pursuant to step 12A, speaker 19 is -turned 18 off by -transmitting "quiet code" to -the PCM buffers 19 200. Quiet code is comprised of a sequence of PCM
signal samples each of which consists of a plurality 21 of logic low signals, (i.e. binary zeroes). The 22 generation of quiet code is discussed in greater 23 cletail below with reference to Figures 3 and I.
2~ In the event the result of the subtraction in s-tep 7A is negative, (i.e., there is no voice 26 energy on the microphone channel), the speaker channel 27 is given a opportunity -to regain control. Firs-tly, 28 -the contents oE regis-ter 208 are sub-tracted from the 29 value PL stored in register 206. The result (PL-NL) is then stored in regis-ter 212. Then, the 31 signal stored in RAMP register 210 is sub-trac-tecl from 32 (PL - NL,) to rernove all traces oE microphone 33 return loss signal from the values stored in line peaX
3~ detec-t register 206. The resu]t of -this subtraction is then comparecl to the -threshold value THL. If the 36 result is negative, then con-trol is not switched to 37 the speaker channel and the speaker is once again ~3~ 3 01 turned off, in s-tep ]2A. If however, the result of 02 -the subtraction in s-tep 13A is positive, -then control 03 is transferred to the speaker channel by setting the 04 bit stored in control flip-flop 216 to a "0", in step 05 14A. fence when the program recycles and returns to 06 step 6, -the "0" stored in control Elip-flop 216 07 indicates that the speaker channel is in control.
OE3 Next, the value s-tored in register 212, 09 (PL-NL), ls multiplied by 16 (ie. shifted bits to the left in accumulator 201), stored in register 11 214, and cornpared in step 8B with -the contents of RAMP
12 register 210. If the resu]t of this comparison is 13 negative, the microphone is turned off in step 12B, by 14 generating and transmit-ting quiet tone to the remo-te party. If however the result oE the comparison is 16 positive, RAMP register 210 is loaded with the value 17 16(PL - NL) stored in register 214.
18 The value 16(PL-NL) is an estimate of 19 the predicted room echo signal amplitude. The signal gain between speaker 19 and microphone 10 including 21 amplifiers 11 and 18 (Figures lA, lB) is typically 22 20db (hence the multiplication by 16 . If the 23 predicted signal amplitude is larger than the value 24 stored in RAMP register 210, then ~ArlP register 210 is set to be equal to 16 (PL-NL). RAMP register 210 26 is continually updated in this way and decremented 27 pursuant to s-tep 5 such that channel switching due to 2~3 echo signals is vir-tually eliminated.
29 Wi-th reference -to step 12B, -the microphone is effectively switched off by transmitting q~liet code 3] to the PCM buffers 200, so as to ensure tha-t signals 32 emanating Erom the spea~cer are not transmitted to the 33 remo-te subscriber via the microphone.
3~ Steps 7B, 13B and 14B correspond to s-teps 7A, 13A and 14A referred -to above. Hence, the steps 36 executed pursuant to the speaker challnel being in 37 control correspond to the s-teps referred to above with 01 respec-t to -the microphone channel being in control, 02 with -the exception -that due to the large (20db) gain 03 between -the speaker 19 and microphone 10, the 04 multiplication factor 16 is used as opposed to 8 or 05 the microphone channel such that suEficient 06 cancella-tion of reverberation is achieved. Also, 07 microphone gain control is not implemented when the 0~ speaker channel is in control. As discussed above, 09 microphone gain control substantially eliminates the transmission of noise from the local subscriber using 11 a loudspeaking telephone, to a remote par-ty, and :L2 consequent]y is unnecessary when the speaker channel 13 is in con-trol.
l In the even-t tha-t the remote party s-tops speaking, and his last spoken words reflect off a wall 16 or other reflector in the acoustical environment of 17 the local subscriber's loudspeaking -telephone, causing 18 echo signals, -the microphone 10 receives these echo 19 signals and applies them to PCM bu-ffers 200 during subsequent samples. Since the RAMP register 210 21 follows the signal peaks on the speaker channel, the 22 value stored in RAMP register 210 will be large.
23 Because the remote party has stopped speaking, PL is 2~ a small value and the subtraction in step 7B yields a negative result. In step 13B, the microphone echo 26 signal stored in RUMP register 210 is subtracted from 27 the value (Pu Nu) s-tored in register 211. The resul-t 2~ of this subtraction is then compared to a -threshold 29 value THU, s-tored in register 209.
Because -the value stored in RAMP register 31 2L0 is large, as described above, the result of the 32 subtraction in s-tep 13B will be nega-tive. A negative 33 result in the subtraction in step 13B indicates that 3~ the line channel i5 to retain control. In other words, if the remote party pauses while speaking, echo 36 signals picked up by the microphone 10 at the 37 subscriber's loudspeaking telephone will not be large 3~3 - 23 -3 ?. r J it l enough -to cause -the microphone channel to regaln 02 control.
03 In the even-t that both parties remain 04 silen-t after the echo signals have diminlshed, PL
05 remains low and subsequent subtractions pursuant to 06 steps 7B and 13B yield negative results The line 07 channel -therefore retains control. However, -the value 08 stored in RAMP register 210 is decremented with every 09 received neyative PCM frame (step 5) and gradually :L0 re-turns to zero. the line channel retains control 11 un-til such time as -the effective mic.rophone signal 12 amplitude (Pu - Nu) stored in register 211, is greater L3 than the sum of the values provided by THU and RAMP.
14 Threshold value THU establishes a threshold level below which control will not inadvertently switch as 16 result of reception of low amplitude signals on the 17 microphone channel, (eg. background spurious noise 18 etc.).
lg By the same principle, if both -the local subscriber and the remote party begin talking 21 simultaneously and equally loudly, the channel which 22 was previously in control will retain control.
23 If, however, the local subscriber begins 24 talking while the remote party has paused, -the current sample of Pu becomes greater than the sum of Nu and 26 threshold value THU, (the value RAMP is decremented to 27 approximately zero after approximately 100 msec.). As 28 a result, -the subtraction in step 13B yields a 29 positive result, and control flip-flop 216 is loaded with a "1" in step 14B (i.e. -the microphone channel 31 talces control). The value (PU-NU) stored in register 32 207, is multiplied by eight and the resul-t is s-tored 33 in register 211. RAMP register 210 is loader with -the 3~ binary value 8(PU - Nu) stored in register 214 in the event the RAMP value is less than 8(PU Nu), pursuant 36 to execution of steps 8A and 9A. The microphone gain 37 control is adjusted in steps 10, llA and llB and -the ;3 01 speaker 19 is turnecl off step 12A).
02 In summary, the digital signal proeessor 03 portion of the loudspeaking telephone detects whieh of 04 -the line or microphone channels was previously in 05 con-trol, yet provides an opportunity for the other 06 channel to regain control by comparing incoming and 07 outgoing envelope signal sample values with 0~ corresponding predetermined threshold and echo 09 values. The line ehannel signal samples are suppressed in the event -the local subscriber is Ll talking, and -the microphone ehannel gain is adjusted 12 in response -to signals thereon. Similarly, the :L3 mierophone signal samples are suppressed in the event 14 the remote party is talking. The digital signal proeessor guards firstly against remote and loeal 16 loudspeaking telephones being simultaneously switehed 17 into their mierophone modes, and secondly against l spurious channel switehing due to eeho signals or in 19 the event neither party is talking or both parties are talking equally loudly.
21 The present invention also compensates for 22 noise generated on the incoming and outgoing channels 23 due to crosstalk, etc. by detecting average noise 24 floor signal sample values associated with the channels and subtracting them from voice signals 26 carried by the channels.
27 Figures 3 and 4 are schema-tic diagrams of 28 data storage and manipulation, and timing and eon-trol 29 eireuitry according to a preferred embodimen-t oE the digital signal processor 14 of -the invention.
3L With reference to Figure 3, incoming and 32 outgoing Kit PCM signal samples PCMi and PCMo are 33 loadecl serially in-to PCM buffers 200 on a positive 34 transition of a system elock signal I, in response to e~ecu-tion of a SHIFT LEFT instruction generated by an 36 instruction ROM 41~ (Figure 4~, discussed in greater 37 de-tail below.

J~3 01 PCM bufEers 200 are connected -to PCM IN
02 and PCM OUT -terminals of bidirec-tional accumulator 03 201. Accumulator 201 is connec-ted -to the Sl inpu-t of 04 ar-thimetic logic unit ALU 204. Storage registers 05 205-214 are connected -to the S2 inpu-t of ALU 204 as 8q described with reference to Figure 2. Enable inputs 08 E, of regis-ters 205-214 are connected to respective 09 outputs Y0-Y5 and Y0-Y2 of -three-to-eight decoders 307 and 304 respec-tively. In a successEul prototype of Ll the inven-tion, accumulator 201 and registers 205-214 l2 each had 14 bit capacity.
L3 Registers 205-214 can be any of a varie-ty l oE s-torage regis-ters such as random access memories.
However, in the preferred embodiment, registers 16 205-214 are shift registers.
17 A QA output o-f accumulator 201 is 18 connected to an I2 input oE PCM bufEers 200. The QA
19 output transmits the least significant 8 bi-ts of a 14 bit value stored -therein to PCM buffers 200, which 21 typically store 8 bit signals.
22 Noise detec-t registers 207 and 208 are 23 parallel loadable from a microprocessor, R, not 24 shown, for performing DTMF tone genera-tion, as described below. Threshold register 209 is parallel 26 loadable for storing threshold values THU, TH2 and Tu, 27 discussed above. Parallel outpu-ts Q0-QF, of RAMP
28 register 210 are used to address a DATA ROM 306.
29 Signals appearing on -the serial ou-tputs S0, of registers 205-214 can be selectively Eed back 31 to respective inputs SI thereof -through gates 308 and 32 3:L0, under con-trol of data bi-ts D3, D4 ancl D5.
33 Before describing operation of -the 34 circuitry illustrated in Figure 3, reEerence will be made -to Figure 4, illustrating -the -timiny and con-trol 36 circui-try 202, discussed wi-th reEerence to Figure 2.
37 A high Erequency oscillator 402 generates 01 an approximately 4.096 MHz signal to a master clock 02 circuit 404. In response to recep-tion of the high 03 frequency signal, master clock circuit 404 generates 04 the aEorementioned clock signal 0 Erom a CLK ou-tput 05 thereof. In addi-tion, master clock circuit 404 06 generates a frame pulse signal FP from -the RESTART
07 output thereoE, which is applied to -frame counter 8~ flip-flop 406 via an inverter 407.
:L0 The Q output of flip-flop 406 is connected ll to the D input thereof, such that the flip-flop :L2 func-tions as a toggle in response to receiving frame 13 pulse s:ignals FP from master clock circui-t 40~. The Q
l output oE flip-flop 406 is connected to jump control circuitry, described in further detail below.
16 Outputs Q0, Ql, Q2 and Q3 of a master 17 coun-ter 408 are connected to the A, B, C and D inputs 18 of a 4-to-16 decoder 410.
19 Outputs Y2, Y4, Y7 and Y13 oE decoder 410 are connected to DECODE inputs of a reset circuit 21 412. Data bits D2, D3, M and D9, in addition -to 22 enable signals SHLAB, SHRB, REGLD and ACCONTB, are 23 applied -to CONTROL inputs of reset circuit 412.
24 Predetermined combinations of signals on the CONTROL
and DECODE inputs of reset circuit 412 cause 26 genera-tion of reset signals from an EN output 27 -thereof. ale EM output is connected to a clear input 28 CLR of master counter 408 and a coun-t enable input CEN
29 of a program counter 41~ via an inverter 416. Also, the frame pulse signal FP generated by master clock 31 ~04 :is applied to -the CONTROL inputs of reset circui-t 32 412.
33 Mas-ter counter 408 counts -Erom 0 to 14 34 (dec1mal), after which the EN output of reset circuit 412 goes momen-tarily low, applying a reset signal to 36 counter 408. Master counter 408 coun-ts -the number of 37 cycles required Eor execution of an instruction s-tored 3~ - 27 -01 in ROM 418. Instructions require varying lengths of 02 time Eor their execution, ranging from between 1 to 14 03 clock cycles. For example, adding two binary samples 04 through ALU 204 requires 14 cycles, and a jump 05 ins-truction requires only 1 cycle. Each time the 06 counter 408 is reset, program counter 414 is enabled 07 for 1 clock cycle, and incremented by 1, thereby 08 addressing -the next loca-tion in ROM 418. The contents 09 of the addressed location in ROM 418 are decoded in :L0 decoders 302, 304 and 312 (in Figure 3), and 420, LI (shown in Figure 4), in order to implement one of L2 seVen different types oE instructions; JUMP, INCDEC, 13 ARITHMETIC, SHIFT LEFT, SHIFT RIGHT, REGISTER LOAD and 14 ACCUMIJLATOR CONTROL.
Various ones of the seven instruction 16 types cause performance of the various operations 17 described with respect to Figure 2 and the flowchart.
18 Master counter 408 is also reset in 19 response to SHLAB, SHRB or ACCONTB signals appearing on the CONTROL inputs of reset circuit 412. For 21 example, mas-ter counter 408, and program counter 414 22 are reset upon receipt of the frame pulse signal FP
23 prom master clock 404. In a successEul embodiment of 24 the invention, the frame pulse signal occurs every 125 microseconds. The ARITHMETIC and INCDEC instructions 26 typically require 14 clock cycles for execution, the 27 SHIFT LEFT and SHIFT RIGHT instructions require prom 1 28 to 8 clock cycles for execution, and the JUMP, 29 ACCUMULATOR CONTROL and REGISTER LOAD instructions require only one cycle for execu-tion.
31 Signals appearing on -the most significan-t 32 clata ou-tputs D6, D7, D8 and D9 of instruc-tion ROM 418 33 are decoded in decoder 420 in order to provide a 34 number of jump control signals. Condi-tional j~np signals Erom outputs Y2, Y4 and Y5 thereoE are 36 inver-ted in inverters 422, 424 and 426 respectively 37 ancl applied to first inputs oE AND gates 428, 430 and 38 - ~8 -01 432, respectively. The SGN signal from accumulator 02 201 (Figure 3) provides an indication of whether a 03 value stored therein is positive or negative, (i.e., 04 the SGN signal is the most significant bit of the 05 value s-tored in accumulator 201). The SGN signal i5 06 applied to a second inpu-t of AN gate 428. A second 07 input of AND ga-te 430 is connected to the Q output of 08 flip-flop 407.
09 Output Y0 of decoder 420 is inverted in an L0 inverter 434 to provide an unconditional jump signal, :Ll for application to a first input of a NOR gate 436.
L2 The outpu-t of inverter 434 and the outputs of AND
L3 gates 428 and 430 are connected to second and third 14 inputs of NOR gate 436, the output of which is connected to a first input of an AND gate 438.
l6 Clock signal 0, data bit D3 and an enable 17 signal CTEN are applied to first, second and third 18 inputs of an AND ga-te 440. The output of AND gate 440 19 is connected to a clock input of control flip-flop 216. The D input of flip-flop 216 is connected to the 21 SGN output of accumulator 201. The Q output of 22 flip-flop 216 is connected to a second input of AND
23 gate 432. The output of AND gate 432 is connected -to 24 a first input of a NOR gate 442, and generates an enable signal denoted as JMPSGN~, for effecting a 26 predetermined one of the JUMP instructions. A second 27 input of NOR gate 442 is connected to ground, and the 28 output thereof is connected to a second inpu-t of AND
gate 438. The output of AND gate 438 is connected -to 31 a PCLD input of program counter 414.
32 When any of the outputs of AND gates 428, 33 430 and 432 or inverter 434 go high, the output of 34 corresponding NOR gates 436 or 442 go low, causing the ou-tput of AND gate 438 to go low. As a result, 36 program counter 414 is loaded wi-th data bits D0 to D6, 37 Erom ROM 418, which represent an interrupt address ~æ-01 location in ROM 418. Hence, program control "jumps"
02 to an interrupt address in response to predetermined 03 signals being applied to AND gates 428, 443, 432 and 04 inverter 434.
05 Output Y3 of decoder a~20 generates a GCBN
06 enable signal for application to the gain control 07 circuitry 16A, discussed with reference to Figure lA, 08 and in greater detail below with reference to Figure 09 5.
Returning to Figure 3, data bits D6, D7 Ll. and D8 are decoded in the previously mentioned three-to-eight decoder 312. Data bit D9 is connected 14 to an enable input EN, of decoder 312, such that one of decoders 312 and 420 (Figure 4) is disabled when g the other one of the decoders is enabled. Output Yl 18 of decoder 312 is connected to the enable input EN of 19 decoder 304. Output Y2 of decoder 312 generates a signal ACCONTB, for efEecting one or the aforemention 21 ACCUMULATOR CONTROL instructions. The ACCONTB signal 22 is inverted in an inverter 314 to provide the 23 aforementioned control enable signal CTEN. The ou-tput 24 of inverter 314 is connected to first inputs of AND
gates 316 and 318. Second inputs of AND gates 316 and 26 318 are connected respectively to the D4 and D5 data 27 outputs of instruction ROM 418 (Figure 4). The output 28 of AND gate 316 goes high, generating a SET SGN
29 signal, in response to the Y2 output of decoder 312 going low and data output D4 of ROM 418 (Figure 4) 31 going high. The SET SGN signal causes the sign bit 32 SGN of a value stored in accumulator 201 -to be set 33 equal -to 1. Similarly, the output of AND gate 318 34 generates a CL,R SGN signal for clearing the sign bit of a value stored in accumulator 20] in response -to 36 the ~2 output of decoder 312 being low and the D5 data 37 output being high.

Lo 01 Decoder 304 is enabled in response to the 02 Yl output of decoder 312 going low. The Yl output of 03 decoder 312 is also connected to a first input of an 04 AND gate 320. The second input of AND gate 320 is 05 connected to the Y6 output of decoder 312. rule output 06 of AND gate 320 generates an ALUB enable signal for 07 application to accumulator 201, and ALU 20~, for 08 implernenting a predetermined one of the ARITHMETIC
09 in6trwctions.
The Y3 output of decoder 312 generates a :Ll REGLDB signal :Eor effecting one of the aforementioned 12 R:EGISTER LOAD instructions which is applied to a 13 register load input LD of threshold register 209, for 14 load.ing the thres'nold values THU, THL and Tu.
The Y4 output of decoder 312 generates a 16 SHRB signal for implementing one of the SHIFT RIGHT
17 instructions, which is applied to first inputs of NOR
18 gates 322 and 32~. A second input of NOR gate 322 is 19 connected to the D5 data output of instruction ROM 418 (Figure 4). Data output D5 of instruc-tion ROM 41~
21 (Figure 4) and the SGN output of accumulator 201 are 22 applied to first and second inputs of a NAND gate 326, 23 the output of which is connected to a second input of 2~ NOR gate 324. The outputs of NOR gates 322 and 324 are connected to first and second inputs of an OR gate 26 328. The output of OR gate 328 generates a SHIFT
27 enable signal, which is applied to a CONTROL input of 28 accumulator 201.
29 Output Y5 of decoder 312 generates a SHLAB
enable signal, which is also applied to -the control 31 input of accu~nulato.r 201.
32 Output Y7 of decoder 3:12 generates an 33 INCDEC enable signal, for incrementing or decrementing 3~ a value applied to ALU 204. The Y7 output of decoder 312 is also connected to a first input of an AND gate 36 330, -the second inpu-t of which is connected to the Y6 37 output of decoder 312. An output of AND gate 330 is 02 connected to an enable input E oF decoder 302.
03 The SHIFT and SHLAB signals enable data 04 stored in accumulator 201 to be shifted to -the right 05 or left as described in greater detail below with 06 reEerence to TABI.ES E and F.
07 The ALUB signal controls arithme-tic 08 opera-tions in ALU 204 and storage o-f the results oE
Og the operations in accumulator 201.
The SET SGN and CLR SGN signals Erom AND
Ll ga-tes 316 and 318 respectively, cause the most L2 signiEicant bit of data stored in accumulator 201 -to :L3 be set -to a "1" or "0" respectively. For example, 14 pursuant to steps 1 and 3 shown in appendix A, an incoming signal is required to be rectified, and in :L6 step 14A the control word is required to be set to a 17 positive value these two requirements are fulfilled 18 by generating and applying the SET SG~ signal to 19 accumulator 201. Also, in step 14B, it is required to negate the control signal stored in flip-flop 216.
21 This is accomplished by generating the CLR SGN signal 22 and applying it to accumulator 201, and subsequently 23 loading the SG~ signal (i.e. zero) into the data input 24 D of flip-flop 216.
Data bits D3, D4 and D5 are connected to 26 control inputs of ALIJ 204 for selection between 27 adding, subtracting, and stream blanking operations of 28 ALU 204, described in detail below with reference to 29 TALES Cl, C2 and C3.
Least significant data bits D0, Dl and D2 3L are decoded in decoders 3~2 and 304 in response to the 32 decoders being enabled. Decoder 302 is enabled in 33 response to receiving a logic low signal Erom -the Y7 34 outpu-t oE decoder 312. The decoded outputs Y0-Y5 of decoder 302 are used to selectively enable shift 37 registers 205 through 214, via enable inputs E

`3 01 thereof.
02 Outputs Y6 and Y7 are connec-ted to El and 03 E2 inputs ox data ROM 306, for selecting between high 04 and low pages of data stored therein.
05 Decoder 304 is enabled in response to the 06 Yl output ox decoder 312 going low. Data bits ~0, l 07 and D2 are decoded in decoder 304 to provide enable 0~ signals for transmission to shift registers 211, 212 09 and 214.
Opera-tion of -the digital signal processor :L1 as a loudspeaking telephone is initiated at the start 12 of each 125 microsecond frame by the Erame pulse signal 13 FP, generated by the master clock circuit 404, which 14 causes -the CLR input of program counter 414 to go high and the EN output of reset circui-t ~12 to go low thus 16 clearing counters 408 and 414 and thereby addressing 17 the first memory location of instruction ROM 41~
18 Serial PCM buffers 200 receive the current 19 line and microphone PCM signal samples, PCMi and PCMo, and store one time slot of each for the remainder of 21 the 125 microsecond frame, during which time the 22 samples are processed. Following this time slot which, 23 in the successful prototype, was approximately 24 microseconds, instructions in ROM 418 are decoded in decoder 312 such that the signal samples stored in PCM
26 buEfers 200 are shiEted, suppressed or otherwise 27 arithmetically manipulated.
2~ PCM words are shifted most significant bit 29 firs-t, but ALU 204 is required to be loaded with -the least significant bit first. Thus, accumulator 201 is 31 made bidirectional such that words are shifted into the 32 PCM IN input -thereof mos-t significant bit first and 33 shifted out the QR output thereof least siynificant bit 3~ first. Similarly, data is shifted in-to the input IN, leas-t signiEican-t bit first and shifted from PCM OUT
36 and QA most significant bit first.
37 In order to rectiEy an incoming signal 3~ - 33 -01 sample stored in accumulator 201, the signal sample is 02 shifted into accumulator 201 and the current 03 instruction data bits D0-D9, generated by ROM 418 are 04 decoded in decoder 312 such that output Y2 goes low, 05 and data bit D5 goes high such tha-t the output of AND
06 gate 318 goes high clearing the most significant bit 07 (sign bit) of -the sample stored in accumulator 201 to 08 zero. This corresponds -to rectifying the input sample, 09 as described above with reEerence to Figure 2.
L0 on order -to se-t or resek control flip-flop ll 2:L6, as described above, the SGN signal from 12 accumulator 201 is applied to the D input -thereof.
L3 Data bits D6-D8 are decoded in decoder 312 such that 14 the Y2 outpu-t thereof goes low. With data bit D3 at a L5 logic high level, and on a rising edge of clock signal 16 0, the output of AND gate 440 goes high, la-tching the 17 value of SGN into control flip-flop 216.
18 In order to load one ox the threshold words 19 THU, THL or Tu into threshold register 209 for subtraction in ALU 204, the data bits D6-D~ generated 21 by instruction ROM 418 are decoded such that output Y3 22 of decoder 312 goes low, thereby applying the REGLDB
23 signal to the LD inpu-t of register 209. In addition, 24 data bits D0-D2 are decoded in decoder 304 such that output Yl thereof goes low in response to being enabled 26 as a result of outputs Y6 and Y7 of decoder 312 being 27 at high logic levels. As a resul-t, the six least 28 significant data bi-ts, D0 to D5 (corresponding to -the 29 threshold word), are loaded into register 209. The threshold word is then subtracted -from -the difference 31 between the peak and noise values, in ALU 204, and the 32 result is stored in accumula-tor 201. Next, the value 33 of RAMP stored in register 210 is subtracted Erom the 34 result in accumulator 201, as discussed with reEerence to steps 13A and B in -the Elowchart.
36 In the event that the resul-t oE a 37 sub-traction in ALU 204 is negative, the SGN output of 3 r 01 accumulator 201 goes high, providing the "condition"
02 for a condi-tional jump to AND gate 4~8, (Figure I).
03 Data bits D6, D7 and D8 from instruc-tion ROM 418 are 04 decoded in decoder 420 such that output Y2 goes low, 05 and program counter 41i is loaded with a predetermined 06 jump address.
07 In steps 9A and 9B described with reference 08 to appendix A and Figure 2, RAMP register 210 is loaded 09 wi-th the values 8(PU-Nu) and 16(PL-NL), respec-tivel~. Mu]tiplication by 8 is performed in :I.L accumulator 201 by shiEting the (PU-NU) value three 12 blts -to the left, and multiplication by 16 is performed 13 by shif-ting four bits to -the left.
lo Signal samples s-tored in predetermined ones of the shift registers 205~214 can be shifted out from 16 the SO output thereof and fed bacX to -the SI input 17 -thereof via transmission gate 308 in response to 18 receiving a logic low enable signal frown an AND gate 19 332. A first input of AND gate 332 is connected to the D5 output of ROM 418, the second input of which is 21 connected to an output of NAND gate 334. Two inputs of 22 NAND gate 334 are connected to the D3 and D4 outputs of 23 instruction ROM 418. A selected one of registers 24 205-214 receives signal samples on the corresponding SI
input thereof, via one of transmission gates 308 or 3]0 26 in response to prede-termined values of da-ta bi-ts D3, D4 27 and D5 applied to A~JD gate 332 and NAND gate 334, as 28 described in greater detail below wi-th reference to 29 descrip-tion of the ARITHMETIC instructions.
As discussed above, da-ta bits D0-D9 31 genera-ted by instruction ROM 418 are decoded to provlde 32 the aforementioned JUMP, INCDEC, ARITHMETIC, SHIFT
33 Ll3E'T, SHIFT RIGHT, REGISTER LOAD and ACCUMULATOR
3~ CONTROL instructions for performing the various operations described with reference to the flowchart in 36 appendix 2.
37 Each of the instruc-tions has a variety of 3~

l forms, depending upon specific values of da-ta bits 02 D0-D9.
03 The JUMP instructions generate signals 04 which ultimately result in the output oE AN gate 438 05 going low, causing data bits D0-D5 of ins-truction ROM
06 418 -to be loaded into the program coun-ter 414. The 07 various JUMP instructions generated in response to 08 decoding data bits D0-D9, are illustrated in TABLE A.
Og 1 0 _____.____ __ _ ____ _ __ l:L TABLE A
12 _ __.__._ _ _ __. ..............
:l3 D9D8D7D6D5D4D3D2DlD0 Description 16 1 0 0 0 JUMP ADDRESS uncondltional jump address 17 1 0 0 1 JUMP ADDRESS not used 18 1 0 1 0 JUMP ADDRESS jump to address if SGN=l 19 1 0 1 1 JUMP ADDRESS enable gain control 1 1 0 0 JUMP ADDRESS jump to address if FRAME
21 is positive 22 1 1 0 1 JUMP ADDRESS jump to address if JMPSGNL=l 23 1 1 1 0 JUMP ADDRESS not used 24 1 1 1 1 JUMP ADDRESS not used 26 where X = don't care _ _ 28 The contents of registers 201 210 are 29 incremented or decremented in response to execution of the INCDEC instruction generated by decoding data bits 31 D0-D9 according to the values shown in TABLE Bl and 32 selec-ting one of the registers or a high or low page o 33 ROM 306 according to the va].ues shown in TABLE B2, as 34 Eollows:

02 TABLE Bl 03_ _ _ ___ 05D9D8D7D6D5D4D3 Description 070 1 1 1 X 0 0 increment selected register by one 08 if SGN=0, else decremen-t by 8 090 1 1 1 X 0 1 increment selected register by two if SGM=0, else decrement by 8 110 1 1 1 0 1 0 increment selectd register by zero 12 if SGN=0, else increment by 1 L30 1 1 1 0 1 :L always increment hy one .l40 1 1 1 1 X 0 increment selected reyister by one if SGN=0, else decrement by 8 160 1 1 1 1 X 1 incremen-t selected register by two 11 if SGN=O, else decrement by 8 18where X= don't 20care _ __ ____ _,_ 22 __ _ _ _ _ _ _ _ 24 _ _ 25D2DlD0 Description 270 0 0 select register 210 280 0 1 select register 209 290 1 0 select register 205 300 1 1 select register 206 311 0 0 select register 207 321 0 1 select register 208 331 1 0 select low page of data ROM 306 351 1 1 select high page of data ROM 306 37Data on the Sl and S2 inpu-ts of ALU 204 can 38 be added together, subtrac-ted, or otherwise 39 arithmetically and logically manipulated in response to generation of the ARIT~IMETIC instruction, which is 41 generated in response to data bits D3-D9 being decoded ~2 as shown in TABLES C]. and C2. Data on the S2 input of 43 AI.U 204 ls provided from a selected one of registers 4~205-214. Registers 205~210 as well as da-ta ROM 306 ~5- 37 -s~3~

01 are selec-ted according to the values of D0-D2 sho~m in 02 TABLE B2 above. Shift registers 211-214 are selected 03 according -to the values shown in TABLE C3 below in the oa, event da ta bit D6 is at a high logic level and bits 05 D7-D9 are a-t logic low levels, according to TABLE: C2 06 below.

rJ 8 0 9 TABLE Cl :1. 0 _ __.__.___.____ _ _.__ __ i2 _5D4D3 Description L30 0 0 add selected register to accumulator, i store in accumulator :L50 0 1 subtract selected regis-ter from 16 accumulator, s-tore i.n accumula-tor :L70 1 0 shift selected register in-to 18 accumulator 190 1 1 shift 2's compliment of selected register into accumulator 211 0 0 add selected register to accumulator, 22 store in register and accumulator 231 0 1 subtract selected register from 24 accumulator, store in register and accumulator 261 1 0 shift accumulator into selected 27 register, save accumulator 281 1 1 decrement accumulator 29 _.

__ . . _. ______ __ __ 34D9D8D7D6 Description _ 360 1 1 0 arithmetic operation on registers 380 0 0 1 arithme-tic operations on register O____. _____._ ________ ____.______._._.___ _.___.___ ~2~3~
o 1 04 .___ _ _ _._ _.~ _ _ __ _ _ _ _ _ _ __ _ _ 06 D2DlD0 Description 07 0 0 0 select register 211 08 0 0 1 select register 212 09 0 1 0 select register 214 1 0 _______,___._ _______ ____ l.L
:1.2 PCM values are shifted between accumulator :l.3 201 and PCM buffers 200 in response to execution of :L4 particular forms of the SHIFT LEFT instruc-tion, as :l5 shown below in TABLE D:
:L6 :L7 TABLE D
i9 ________________ ________ 20D9~8D7D6D5D4D3D2 Description 220 1 0 1 1 0 1 0 shift from accumulator to PCM
23 buffers (speaker channel) 240 1 0 1 0 0 1 0 clear PCM buffers (speaker channel) 260 1 0 1 1 1 1 0 shift from accumulator to PCM
27 buffers (microphone channel) 280 1 0 1 0 1 1 0 clear PCM buffers (microphone 29 channel) 300 1 0 1 X 0 0 0 shift from PCM buffers 31 (microphone channel) to 32 accumulator 330 1 0 1 X 1 0 0 shift from PCM buffers (line 34 channel) -to accumulator 350 1 0 1 X X X 1 shift conten-ts of accumulator 36 one bit left 37where X-- don't 38care 39__ _ ___ _ _ _ _ __ _ _ _ _ 41 Da-ta s-tored in accumulator 201 can be ~2 cond:itionally or unconditionally shifted 1, 3 or 5 43 bits to the r.ight -therein in response -to execution of ~4 -the SHIFT RIGHT ins-truc-tion, as shown below with reference -to TABLE E:

~6 39 -~,~33~

o 1 0~ D9D8D7D6D5D4D3D2 Description 06 0 1 0 0 0 0 0 1 shift con-tents of accumulator 07 1 bit right 08 0 1 0 0 0 0 1 0 shift contents of accumulator 09 3 bits right 0 1 0 0 0 1 0 0 shift contents of accumulator l:L 5 bits righ-t 12 0 1 0 0 L 0 0 1 shift contents of accumulator 13 1 bit right, if SGN=l 14 0 :L 0 O 1 0 1 0 shift contents of accumulator 3 bits right, if SGN=l 16 0 1 0 0 1 1 0 0 shift contents of accumulator 18 _ 5 bits right, if SGN=l Threshold register 209 is loaded wi-th a 2:L threshold value defined by data bits D0-D4 in response 22 to data bits D8 and D9 being at logic low levels and 23 data bits D6 and D7 being at logic high levels, 24 (REGISTER LOAD instruction).
The most significant bit oE a value stored 26 in accumulator 201 (the sign bit) can be cleared or 27 set, as described above, in response to execution of 28 the ACCUMULATOR CONTROL instruction, as illustrated 29 below with reverence to TABLE F:

33~ _ . _. _ _ _ _ 34D9D8D7D6D5D4D3D2DlD0 Description 360 0 1 0 1 0 0 0 0 0 clear sign bit in 37 accurnulator 380 0 1 0 0 1 0 0 0 0 set sign bit in 39 accumulator 400 0 1 0 0 0 1 0 0 0 latch sign bi-t into 41 flip-flop 216 420 0 l 0 0 0 0 1 0 0 latch sign bit into yain 43 Elip-Elop 216 44_ ______ _ _____ _ _~
~5 46 The circuitry described herein rnay be 47 used to generate DTMF and ringing tones under control 48 o:E an external microprocessor, not shown.

49 - ~0 -01 l'he low order page of data ROM 306 (Figure 02 3) preferably contains a 32 word sinewave loo~up table 03 and the high order page preEerably contains a 0~ linear-to- law conversion table Data ROM 306 is 05 addressed by RAMP register 210 in response -to being 06 enabled by decoder 302. QA-QF outputs of register 210 07 provide address values :Eor addressing particular 0~ loca-t:ions in e:ither -the high or low pages oE data TOM
09 306. A logic high signal applied to -the El input oE
ROM 306 enables the ROM For reading the low page data, ll and a logic high signal on the E2 input enables ROM
12 306 for reading the high page data.
13 DTMF tones are genera-ted by summing two l six bit digitized sinewaves and converting the sum to an 8-bit law encoded value. The contents of RASP
16 register 210, which are applied to address inputs of 17 data ROM 306, are incremented by a tone coefficient 18 value corresponding to a predetermined phase increment 19 ror generating a sinewave at a predetermined Erequency. In order to generate a dual tone, the 21 contents ox register 210 are saved and a second 22 address is loaded into register 210 and subsequently 23 updated by a second tone coeE~icient value generated 2~ by the external microprocesscr.
Microphone envelope detect register 205 26 s-tores -the first value for addressing -the ROM table to 27 generate the first sample oE the Eirst sinewave at -the 28 aforementioned predetermined Erequency, and line 29 envelope detect register 206 stores a second value :Eor addressing the ROM 306 to generate the Eirst sample of 31 the second sinewave havi.ng higher Erequency.
32 M:icrophone noise detect register 207 is loaded with a 33 Eirst tone coe:Eficient from the external 3~ microprocessor. Line noise detect register 208 is loaded wi.t.h a second -tone coeEicient in a like 36 manner.
37 In operation, the contents oE register 205 01 are aclded to the firs-t tone coefficient stored in 02 register 207. The result of this addition is shifted 03 into register 210 for addressing the data ROM 306, 0~ which is enabled in response to a logic high signal 05 applied to the El input thereof. The contents of 06 register 210 are then stored in register 205. Data 07 corresponding to the first sample ox the first 08 sinewave i5 then shifted from the S0 outpu-t of ROM 306 09 into the S2 input of ALU 204, and therefrom into L0 accumulator 201. The sample is then shifted from the ll accurnulator 201, through transmission gate 310 into 12 register 214 in response to at least one of data bits 13 D3 or D4 being low and D5 being high.
l Next, the contents of register 206 are added to the second tone coefficient stored in 16 register 208. The result of this addition is shifted 17 into register 210 for addressing a second location in 18 the data ROM 306. The contents of register 210 are 19 then stored in register 206.
Data from ROM 306, corresponding to the 21 first sample of the second (higher rrequency) 22 sinewave, is shifted therefrom into accumulator 201 23 and added to the sample stored in register 21~. The 24 resulting linear sum is shifted into regis-ter 210.
Next, the high order page of data stored in ROM 306 is 26 enabled in response to a logic high signal being 27 applied -to the E2 input thereof. The sum of the two 28 sinewave samples stored in register 210 is used -to 29 address a location in the linear to law conversion -table stored in data ROM 306. The converted law 31 sinewave sample from data ROM 306 is -then shif-ted out 32 to PCM buffers 200, via ALU 20~ and accumulator 201.
33 The updated con-tents of registers 205 and 3~ 206 are -then addecl to the tone coefficien-ts s-tored in registers 207 and 208 as describecl above and the 36 ent:ire process is repeated 37 In a successful prototype of the 3~ - 42 -3~~

01 invention, the mos-t significant 6 bits of -the values 02 stored in registers 205 and 206 were used or 03 addressing ROM 306 through register 210. The tone 04 coefEicients stored in registers 207 and 208 were 05 required to be bits wide. The resulting DTMF tone 06 was within 1.5% o-f the desired frequency, and tones 07 having :Erequencies between 500 Hz and 1633 Hz were 08 generated.
09 the digital signal processor can also be used as a tone ringer. In a successful embodiment of :l.l the inven-tion the tone ringing signal was a square 12 wave, :Erequency shifted between 500 Hz and 364 Hz at a 13 rate of 16 Fez. The square wave is generated by l loading noise detect registers 207 and 208 from the microprocessor with values representing the number o 16 samples in one half periods of the 500 Hz and 36~ Hz 17 square waves. Register 205 is loaded with the 18 contents of register 207 and then used as a down 19 counter. Register 206 contains a value representing -the amplitude of the square wave.
21 In operation, the contents of the 22 "counter" register 205 are decremented in ALU 204 and 23 loaded in-to accumulator 201 in order to determine 24 whether -the sign bit has been set to whether -the contents oE the counter register have been decremented 26 to 0). The decremen-ted value is then reloaded into 27 the "counter" register. Next, the contents of 2~ register 206, representing the amp:Litude o:~ the square 29 wave, are loaded into PC~ buffers 200 and transmitted to the speaker through codec 12 and amplifier 1~3 3:L (Figures l and lB). The process is repeated until 32 the contents of the counter regi.ster 205 reach 0, in 33 response to which the sign bit of -the ampli.-tude 3~ reg.i.ster 206 is inverted to a negative value representing the amplitude of -the other halE cycle o 36 the square wave. The counter register 205 is again 37 preloaded w:ith the contents o:~ register 207 and the 3~ - ~3 -01 process is repeated.
02 As an alternative, the square wave tone 03 ringer ampli-tude can be made to decay by decre~enting 04 the contents of the amplitude register 206 with every 05 repetition of the above described process.
06 Register 205 is alternately loaded with a 07 eira-t prese-t value stored in register 207 08 (corresponding -to -the 500 Hz frequency) and a smaller Og preset value (corresponding to the 36~ Hz frequency) 1.0 storecl in register 208 at a rate of preferably 16 Hz.
l.l In a successful proto-type of the invention 12 the first preset value was OB hex, and the second :L3 preset value was 08 hex.
l Tone ringing signal samples an DTMF
signal samples are applied from the counter register 16 205 or rom ROM 306 respectively, to ALU 204 and 17 applied -therefrom -to accumulator 201 for application 18 to PCM buffers 200. The signal samples are then 19 applied to on of data lines 15 or 17 (Figures lA and lB) for application to the codec 21 or speaker 19 21 associated with the local loudspeaking telephone.
22 Figure 5 is a block schematic illustration 23 of the gain control circuitry 16A discussed above with 2~ reverence to Figure lA. Referring again to the flow chart in Appendix A, and in particular to step 10, in 26 the event the result of the subtraction PU-Nu-Tu is 27 negative, the SGN output of accumulator 201 goes to a 28 logic high level. Consequently, a logic high level 29 s:i.gnal appears on the D input of a gain control El.ip~flop 500. Data is clocked into flip-flop 500 in 3:L response to the CTEN signal generated by inver-ter 31 32 (Figure 3), -the D2 ou-tput of ins-tLuction ROM l and 33 -the clock signal 0, each going high. The CTEN signal 3~ is applied to a Eirst i.nput of AND gate 502, the seconcl .input of which is connected to the D2 ou-tput of 36 ins-truction ROM ~18. The output of END gate 502 is 37 _ _ 01 connected to a first input of AND gate 504, a second 02 input of which is connected to clock signal I. The 03 output of AND gate 504 is connected to the clock input 04 oE flip-flop 500.
05The Q output of flip-flop 500 goes high in 06 response to the logic high SGN signal being clocked 07 therein. The Q output of flip-flop 500 is connected 08 -to a first input of AND gate 506 and first inputs of 1Osta-te feedback circuits 508 and 510, respectively.
11The Q output of flip-flop 500 is connected 12 to a Eirst input of AND gate 512, and to second inputs L3o~ s-ta-te feedback circuits 508 and 510 respectively.
14Second inputs of AND gates 506 and 512 are connected to up and down clock signal sources UCLK and :L6 DCLK respectively. The up and down clock signal :L7 sources UCLK and DCLK are not shown, but are well 18 known counter circuits typically connected to RESTART
19 output of master clock circuit 404 IFigure 4). The UCLK signal source generates a positive clock pulse 21 once every 2048 frames, and the DCLK source generates 22 a clock pulse each 256 frames. Thus, the output of 23AND gate 506 goes high every 256 milliseconds in the 24 event the SGN output of accumulator 201 is at a logic high level, and the output of AND ga-te 512 goes to a 26 logic high level every 32 milliseconds in the event 27 the SGN output of accumulator 201 is at a logic low 28 level.
29The outputs oE AND gates 506 and 512 are connected to Eirst and second inputs of NOR gate 514, 31 the output of which is connected to clock inputs of fli.p-~lops 516 and 518.
34The Q output of flip-flop 516, and the Q
output of flip-flop 518 are connected to third inputs 36of sta-te feedback circui-ts 508 and 510 respectively.
37 The Q output of Elip-flop 516 is also connected to a ~33~

nl four-th input of state feedback circui-t 510, and the Q
02 outpu-t of ~lip~flop 518 is connected Eurther to a 03 Eourth input of s-tate circuit 508. The outputs of 04 state circui-ts 508 and 510 are connected to the D
05 inputs of flip-flops 516 and 518 respectively.
06 The Q outputs of flip-flops 516 and 518 07 genera-te enable signals denoted as GMS and GLS
08 respectively, for application to state controller 09 circuit 520. 'the enable signal GCEN generated by the :L0 ~3 output of cdecoder 420 (Figure I) is applied to an LI enable input EN of sta-te controller circui-t 520.
l Flip-flops 516 and 518 in conjunction with :l3 corresponding s-tate feedback circuits 508 and 510 l Eunction as the aEorementioned 2 bit up/down gain con-trol counter. Assuming -tha-t the flip-flops are 16 initially reset, and gate 506 genera-tes a logic high 17 signal (i.e. the result of the subtraction in step 10 18 yielded a negative value), the Q output of flip flop 19 518 goes to a logic high level (i.e. GLS equals 1), and the Q output of flip-flop 516 remains at a low 21 logic level (i.e. GMS equals 0). In the event 125 22 milliseconds elapse, and the results of the 23 subtraction in step 10 is still negative, -the 2 bit 24 counter circuit is incremented in response AND gate 506 generating a further logic high signal, such -that 26 GMS equals 1 and GLS equals 0. However, in the event 27 32 milliseconds elapses and the subtraction in step 10 28 ylelds a positive result, a logic low signal is 29 latched into filp-flop 500, causing AND gate 512 to generate a logic high signal which in turn causes the 31 coun-ter circui.t to be decremented.
32 Sta-te feedback circuits 508 and 510 33 operate to ensure that in -the event that both Q
3~ outputs of flip-flops 516 and 51~ are at logic high Levels and ga-te 506 generates a Eurther logic high 36 slgnal, the Q outpu-ts of flip-flops 516 and 518 remain 37 the same (i.e. the counter does no-t "roll over"
38 - ~6 -01 causing GMS and GLS to become bo-th equal to 0).
02 Likewlse, in the event GMS and GLS are each equal to O
03 and gate 512 generates a logic high signal, the 0~ counter is not further decremented.
05 PCM signal samples are received from PCM
06 buffers 200 via data line 15 and applied to a law 07 to linear converter 522 under con-trol of s-tate 0~ controller circuit 520. The chord and step bits are 09 separated in converter 522, and prede-termined ones of LO the step bits are shiEted a prede-termined number of Ll bits to the leEt in order to genera-te a 13 bit linear l2 signal. The linear signal is added to an ofset value 13 (typically 33 decimal) in order to compensate for a l zero crossing conversion offset, in a well known manner. The 13 bit linear signal is received from 16 converter 522 by a serial shift circuit 524, for 17 shifting -the linear signal sample 1 bit to the right 18 (corresponding to an at-tenuation ox -6db), or one bi-t 19 to the left (corresponding to an increase in gain of +6db).
21 The shifted linear signal sample is 22 reapplied to converter 522, or reconversion to law 23 encoded PCM format, under con-trol of state controller 2~ circuit 520.
Hence, -the signal values GMS and GLS are 26 applied to state con-troller circuit 520 in order that 27 linear signal samples applied to serial shift circuit 28 52~ are shiEted 1 bit -to the righ-t or lef-t therein.
29 TABLE G illustrates the values of GMS and GLS, and -the corresponding amount of at-tenuation of 3l the microphone channel signal.
32 _ _____ ____ _ _ _ _____ _____ _ _ _ ___ _ 3~ _ ________ ___ _ _ _ _________ ____ 36 GMS GLS_ r TENUATION
37 O O Odb 38 O 1 - 6db 39 1 0 -12db ~0 1 1 -18db 1 _ .. ___ __ _ _______ _ ~2 - ~7 -~,~3~

01 In summary, the invention is a digi-tal 02 loudspeaking telephone employing a digital signal 03 processor. The dlgital signal processor is capable of 04 implemen-ting various loudspeaking telephone functions;
05 such as channel switching, automatic gain con-trol, ~6 echo cancellation, ringing tone and DTMF tone 07 generation. In the preferred embodimen-t of the 0~ invention an incoming signal from a remote party is 09 suppressed while the local subscriber is speaking, and L0 outyolng signal samples received by the microphone are ]1. gain adjus-ted or transmission to the remote party 12 while he is speaking. In the event both parties are 13 talking simultaneously, the channel which was lA previously in control maintains control, such -tha-t no spurious switching occurs.
16 Persons skilled in the art understanding 17 this invention may now conceive of other embodiments 18 or variations, using the principles described herein.
19 For example, any sui-table digital signal processor may be used, or the incoming and outgoing 21 signals can be digi-tally compressed (i.e. by shifting 22 PCM chord bits a predetermined number of bits to the 23 right), as opposed to being suppressed (i.e.
2~ generating quiet code).
Also, whereas codecs 12 in Figure lA, and 26 21 in Figure lB were described for performing 27 analog-to-digital and digital-to-analog conversion of 2~ audio PCM signal, any suitable A/D or D/A system may 29 be utilized.
All these and o-ther variations are 3L considerecl to be within the sphere and scope o-E this 32 inven-tion, as defined in the clairns appended hereto.

33 - ~8 -

Claims (18)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. A digital loudspeaking telephone for connection to incoming and outgoing unidirectional data lines carrying incoming and outgoing digital signals corresponding to incoming and outgoing analog signals respectively, comprising:
(a) a microphone for transmitting said outgoing analog signal, (b) a speaker for receiving said incoming analog signal, (c) a codec connected to the microphone and speaker, for receiving the outgoing analog signal and generating said outgoing digital signal in response thereto, and for receiving said incoming digital signal and generating the incoming analog signal in response thereto, and (d) a digital signal processor connected to said incoming and outgoing data lines and said codec, for receiving and detecting the relative magnitudes of said incoming and outgoing digital signals and in response suppressing one of said incoming and outgoing digital signals having lesser magnitude relative to the other of said incoming and outgoing digital signals.

2. A loudspeaking telephone as defined in claim 1, wherein said digital signal processor is further comprised of:
(a) means for detecting incoming and outgoing envelopes of said incoming and outgoing digital signals and generating incoming and outgoing digital envelope signals in response thereto, and (b) means for comparing the magnitudes of said incoming and outgoing digital envelope signals and suppressing the incoming digital signal in the event the magnitude of said outgoing digital envelope signal is greater than the magnitude of said incoming digital envelope signal, and suppressing the outgoing digital signal in the event the magnitude of said incoming digital envelope signal is greater than the magnitude of said outgoing digital envelope signal.

3. A loudspeaking telephone as defined in claim 2, wherein said digital signal processor is further comprised of:
(a) means for detecting incoming and outgoing noise floor levels of said incoming and outgoing digital signals and generating incoming and outgoing digital noise floor signals in response thereto, and (b) means for comparing said incoming and outgoing digital envelope and noise floor signals and suppressing the incoming digital signal in the event the difference between the magnitudes of the outgoing digital envelope and noise floor signals is greater than the difference between the magnitudes of the incoming digital envelope and noise floor signals, and suppressing the outgoing digital signal in the event the difference between the magnitudes of the outgoing digital envelope and noise floor signals is less than the difference between the magnitudes of the incoming digital envelope and noise floor signals.

4. A digital loudspeaking telephone as defined in claim 3, wherein said digital signal processor is comprised of a plurality of storage registers for storing said envelope signals and digital noise floor signals.

5. A digital loudspeaking telephone as defined in claim 4, wherein said digital signal processor is further comprised of buffer circuitry connected to said incoming and outgoing data lines for receiving said incoming and outgoing digital signals.

6. A digital loudspeaking telephone as defined in claim 5, wherein said digital signal processor is further comprised of accumulator means connected to said buffer circuitry for suppressing said digital signals.

7. A digital loudspeaking telephone as defined in claim 6, wherein said digital signal processor is further comprised of arithmetic logic unit means connected to said storage registers, for comparing said envelope signals and noise floor signals.

8. A digital loudspeaking telephone as defined in claim 5, 6 or 7, further including automatic gain control means connected to said buffer circuitry for gradually attenuating said outgoing digital signal in the event said incoming digital signal is suppressed and said outgoing digital envelope signal is less than the difference in magnitude between said outgoing noise floor signal and a predetermined threshold signal; and gradually increasing the gain of said outgoing digital signal in the event said incoming digital signal is suppressed and said outgoing digital envelope signal is greater than said difference in magnitude between the outgoing noise floor signal and said threshold signal.

9. A digital loudspeaking telephone as defined in claim 1, further comprised of a control circuit for generating predetermined control signals.

10. A digital loudspeaking telephone as defined in claim 9, further including tone generator means, connected to said control circuit and predetermined ones of said storage registers for receiving predetermined ones of said control signals and generating one of either ringing or dialling tones for transmission to said outgoing data line in response thereto.

11. A digital loudspeaking telephone for connection to incoming and outgoing unidirectional data lines carrying incoming and outgoing digital signals respectively, comprising:
(a) a microphone for transmitting an outgoing analog signal, (b) a speaker for receiving an incoming analog signal, (c) a codec connected to the microphone and speaker for receiving the outgoing analog signal and generating said outgoing digital signal in response thereto, and for receiving said incoming digital signal and generating the incoming analog signal in response thereto, and (d) digital signal processor means for:
(i) detecting incoming and outgoing envelopes of said incoming and outgoing digital signals, and generating incoming and outgoing digital envelope signals in response thereto, (ii) detecting incoming and outgoing noise levels of said incoming and outgoing digital signals, and generating incoming and outgoing digital noise floor signals in response thereto, (iii) detecting the amount of echo signal in a predetermined one of said incoming or outgoing signals and generating a digital ramp signal indicative thereof, in response thereto, (iv) generating incoming and outgoing digital threshold signals, (v) detecting which of said incoming and outgoing digital signals has been previously suppressed relative to the other and selecting step (vi) in the event the incoming digital signal has been previously suppressed and selecting (vii) in the event the outgoing digital signal has been previously suppressed, in which step (vi) is comprised of comparing said outgoing envelope signal with said outgoing noise floor signal and either selecting (A) in the event said outgoing envelope signal is greater than said outgoing noise floor signal or else selecting (B), in which (A) is a step comprised of suppressing said incoming digital signal, (B) is a step comprised of summing said incoming noise floor signal, said incoming threshold signal and said ramp signal and generating a first sum signal in response thereto, and comparing said incoming envelope signal with said sum signal and either suppressing said incoming digital signal in the event said incoming envelope signal is less than said sum signal or selecting (C), in which (C) is a step comprised of suppressing said outgoing digital signal, and step (vi) is comprised of comparing said incoming envelope signal with said incoming noise floor signal and either suppressing said outgoing digital signal in the event said incoming envelope signal is greater than said incoming noise floor signal or selecting (D), in which (D) is a step comprised of summing said outgoing noise floor signal, said outgoing threshold signal and said ramp signal and generating a second sum signal in response thereto, and comparing said outgoing envelope signal with said second sum signal and either selecting (C) in the event said outgoing envelope signal is less than said second sum signal, or else selecting (A).

12. A digital loudspeaking telephone as defined in claim 11, wherein said digital signal processor means is further comprised of:
(a) timing and control circuitry for generating control signals, and (b) data storage and manipulation circuitry connected to said timing and control circuitry and said incoming and outgoing data lines, for receiving said control signals and said incoming and outgoing digital signals, generating and storing said envelope signals, noise floor signals, ramp signals and threshold signals, generating said sum signals and suppressing said incoming and outgoing signals in response to receiving said control signals.

13. A digital loudspeaking telephone as defined in claim 12, wherein said data storage and manipulation circuitry is further comprised of:
(a) a plurality of shift registers for storing said envelope signals, noise floor signals, ramp signals and threshold signals, (b) an arithmetic logic unit connected to said shift registers for generating said sum signals in response to receiving first predetermined ones of said control signals from said timing and control circuitry, (c) a plurality of decoder circuits connected to said shift registers, for enabling predetermined ones of said shift registers in response to receiving second predetermined ones of said control signals from said timing and control circuitry, and (d) buffer circuitry connected to said incoming and outgoing data lines, for temporarily storing said incoming and outgoing digital signals, and suppressing one of said incoming or outgoing signals in response to receiving third predetermined ones of said control signals from said timing and control circuitry.

14. A digital loudspeaking telephone as defined in claim 13, wherein said timing and control circuitry is comprised of:
(a) a clock circuit for generating synchronizing clock signals, (b) a first counter connected to said clock circuit for receiving said clock signals and further predetermined ones of said control signals, and generating count signals in response thereto, c) a reset circuit connected to said counter for receiving said count signals, and generating count enable signals in response thereto, (d) a program counter connected to said reset circuit, for receiving said count enable signals, and generating program count signals in response thereto, and (e) a memory circuit connected to said program counter, for receiving said program count signals and generating said control signals in response thereto.

15. A digital loudspeaking telephone as defined in claim 12, 13 or 14, wherein said digital signal processor means further includes tone generator means for receiving additional control signals, and generating one of either ringing or dialling tones for transmission to said outgoing data line in response thereto.

16. In a loudspeaking telephone, a method of suppressing one of an incoming or outgoing signal on incoming and outgoing data lines respectively, comprising the steps of:
(a) receiving current incoming and outgoing signal samples of said incoming and outgoing signals from said data lines, and generating incoming and outgoing envelope signal samples and incoming and outgoing noise floor signal samples in response thereto, (b) detecting which of a previous one of said incoming or outgoing signals has been suppressed relative to the other, and (c) comparing one of said incoming or outgoing noise floor samples to a corresponding one of said incoming or outgoing envelope signal samples in the event the other one of said previous incoming or outgoing signals was suppressed, and suppressing the other one of the current incoming or outgoing signal samples in the event said corresponding one of the envelope signal sample is greater than the corresponding noise floor signal sample.

17. A method as defined in claim 16, further including the steps of:
(a) generating a ramp signal sample indicative of the instantaneous level of echo signals in said other one of the incoming or outgoing signal samples, (b) generating incoming and outgoing predetermined threshold signal samples, (c) summing the other one of said noise floor signal sample and threshold signal sample, (d) comparing the other one of said incoming or outgoing envelope signal samples with said sum signal sample, and (e) suppressing said other one of the current incoming or outgoing signal samples in the event said other one of the envelope signal samples is less than said sum signal sample, and suppressing the corresponding one of the incoming or outgoing signal samples in the event said other one of the envelope signal samples is greater than said sum signal sample.

18. A method as defined in claim 16 or 17, further including the steps of generating one of either ringing or DTMF tones and applying said generated tones to said incoming line.